PanDM

PanDM performs joint statistical modeling of differential DNA methylation to detect and characterize pan-cancer and cancer-type-specific differentially methylated (DM) CpG sites across multiple cancer types.

Key Features:

• Joint Statistical Modeling: Employs a rigorous statistical model to jointly characterize DM patterns across diverse cancer types, leveraging hidden correlations to enhance statistical power for detecting differential methylation sites.
• Integration with Existing Methods: Operates on summary statistics from individual DM analyses of each cancer type, enabling integration with existing DM calling methods without requiring raw data reprocessing.
• Methylation Site Clustering and Detection: Clusters methylation sites and performs differential methylation detection to identify both common and cancer-type-specific DM patterns across multiple cancers.
• Pan-cancer Pattern Discovery: Discovers pan-cancer DM patterns from datasets such as The Cancer Genome Atlas (TCGA), identifying 37 distinct pan-cancer DM patterns in 12 cancer methylomes to group cancer types by shared methylation characteristics.
• Biological Insights: Uses ontology- and pathway-enrichment analyses to interpret the biological significance of identified DM patterns, including cancer-type-specific etiologies, pathogenesis, and shared environmental risk factors.
• Novel Detection Capabilities: Identifies specific DM CpG sites that are often missed by conventional methods, providing a more comprehensive view of aberrant methylation landscapes across cancers.

Scientific Applications:

• Pan-cancer Research: Provides a systematic framework to investigate aberrant methylation patterns across multiple cancer types and to compare commonalities and differences in tumorigenesis.
• Functional Genomics: Can be extended beyond DNA methylation to other functional genomic profiles such as transcriptomes for broader pan-cancer analyses.

Methodology:

Accepts summary statistics from separate DM analyses as input, applies joint statistical modeling to leverage correlations across cancer types, clusters methylation sites and detects differential methylation, performs ontology- and pathway-enrichment analyses, and is implemented in R.